flightdeck
llms.txt30 starsChanged yesterdayApache-2.0View on GitHub
Flightdeck is a self-hosted observability and control plane for production and coding agents. Every LLM call, MCP event and tool call an agent makes streams to a per-agent timeline and a live fleet-wide feed. Operators set token budgets, MCP server allow/block rules and live directives on production agents (the Claude Code plugin enforces the MCP rules only). Apache 2.0. Production agents add the flightdeck-sensor Python package (pip install flightdeck-sensor, then init() and patch() in the entrypoint); it intercepts the Anthropic,…
- Pipes a download into a shell
- Installs packages
In this repository3 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- dashboard/CLAUDE.mdCLAUDE.md
- sensor/CLAUDE.mdCLAUDE.md
The file
Files
# Flightdeck > Flightdeck is a self-hosted observability and control plane for production and coding agents. Every LLM call, MCP event and tool call an agent makes streams to a per-agent timeline and a live fleet-wide feed. Operators set token budgets, MCP server allow/block rules and live directives on production agents (the Claude Code plugin enforces the MCP rules only). Apache 2.0. Production agents add the `flightdeck-sensor` Python package (`pip install flightdeck-sensor`, then `init()` and `patch()` in the entrypoint); it intercepts the Anthropic, OpenAI and litellm SDKs and the frameworks built on them (LangChain, LangGraph, LlamaIndex, CrewAI), and the official `mcp` client SDK, with no other code changes. Coding agents attach through the Claude Code plugin in the same repository: `flightdeck connect claude-code` installs it and points it at the Flightdeck running on this machine, on whichever port it was started (the install scripts offer to when they find Claude Code; `flightdeck connect` lists the agents the binary can connect; it needs a recent Claude Code and, with an older one, changes nothing and says to run `claude update`), or inside Claude Code `/plugin marketplace add flightdeckhq/flightdeck`, then `/plugin install flightdeck@flightdeck-plugins`. Flightdeck is one `flightdeck` binary (the ingestion API, the workers, the query API, the message bus and the dashboard in one process) on SQLite by default, Postgres as an option. It installs with one command per platform (`curl -fsSL https://flightdeck-hq.com/install.sh | sh` on Linux and macOS, `irm https://flightdeck-hq.com/install.ps1 | iex` in PowerShell on Windows), then runs as `flightdeck` on `http://localhost:4000` (`--port` for another port, `--public` to let other machines reach it, which lets whoever can reach it read its dashboard token: open it only on a network you trust, or put it behind a proxy that authenticates); Homebrew, `pip install flightdeck-cli` and the `flightdeckhq/flightdeck` Docker image install the same binary. The same image runs under Docker Compose and the Helm chart (a StatefulSet: one pod on SQLite by default, several on Postgres with a shared NATS). Flightdeck is not a proxy or LLM gateway (calls go straight to the provider), not multi-tenant SaaS, and does not store prompt content from production agents unless capture is turned on (the Python sensor defaults to `capture_prompts=False`). The Claude Code plugin captures prompts and sanitised tool inputs by default, since you are observing your own session; `FLIGHTDECK_CAPTURE_PROMPTS=false` turns that off. ## Docs - [README](https://github.com/flightdeckhq/flightdeck/blob/main/README.md): install, connecting agents, what you get, coverage in short, self-hosting in short - [Coverage](https://github.com/flightdeckhq/flightdeck/blob/main/guides/coverage.md): every intercepted LLM SDK method, the framework matrices, coding agents, sub-agents, MCP, what is not planned - [Features](https://github.com/flightdeckhq/flightdeck/blob/main/guides/features.md): each dashboard page and control in full - [Self-hosting](https://github.com/flightdeckhq/flightdeck/blob/main/guides/self-hosting.md): Docker Compose, the Helm chart and its values, upgrade notes, backups - [Windows](https://github.com/flightdeckhq/flightdeck/blob/main/guides/windows.md): the PowerShell installer, running Flightdeck, its data, upgrading and removing it, agents from PowerShell - [OpenTelemetry into Flightdeck](https://github.com/flightdeckhq/flightdeck/blob/main/guides/opentelemetry.md): Claude Code's own telemetry and the agent frameworks' GenAI spans - [MCP Protection Policy](https://github.com/flightdeckhq/flightdeck/blob/main/guides/mcp-protection-policy.md): scopes, resolution, configuration and enforcement - [Architecture](https://github.com/flightdeckhq/flightdeck/blob/main/ARCHITECTURE.md): components, event types and payloads, APIs, analytics, content capture - [Claude Code plugin](https://github.com/flightdeckhq/flightdeck/blob/main/plugin/README.md): what the plugin reports for each hook, configuration, privacy - [Python sensor](https://github.com/flightdeckhq/flightdeck/blob/main/sensor/README.md): the `flightdeck-sensor` package, identity, `patch()` vs `wrap()`, environment variables, limitations ## Project - [Releases](https://github.com/flightdeckhq/flightdeck/releases): release notes and artifacts - [Changelog](https://github.com/flightdeckhq/flightdeck/blob/main/CHANGELOG.md) - [Security policy](https://github.com/flightdeckhq/flightdeck/blob/main/SECURITY.md) - [Contributing](https://github.com/flightdeckhq/flightdeck/blob/main/CONTRIBUTING.md) ## Optional - [License (Apache 2.0)](https://github.com/flightdeckhq/flightdeck/blob/main/LICENSE) - [Source](https://github.com/flightdeckhq/flightdeck)
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

